Background 458 patients with rheumatoid arthritis (RA) and unsatisfactory response to DMARDs alone and/or TNFα blocking agents were treated with the IL-1 receptor antagonist anakinra. The initial analysis showed no difference between TNFα blocker non-responders and TNFα blocker naïve RA patients. (ref 1).
Objectives In order to identify factors that could predict response to anakinra treatment we performed a post hoc analysis of the original study data.
Methods Original study data including demographic parameters, concomitant diseases and other anti-rheumatic treatment were subject to multivariate statistical analysis to identify independent factors that impact on the DAS score difference and EULAR response after one year of anakinra treatment. Cohort analyses of patients leaving the study were performed to identify characteristics predictive for study dropout. Data from patients leaving the study prematurely were imputed based on the LOCF (last observation carried forward) methodology.
Results Patients leaving the study prematurely (non-completers) were characterised by less DAS improvement and higher age than completers. After three months of treatment 0.5 units less DAS improvement increased dropout risk by 24% (p<0.001) and one year of additional age increased dropout risk by 3% (p=0.02). Adverse events did not increase the risk of dropout. As patients with poorer response (less improvement of DAS score) had a higher rate of dropout it was important to use LOCF imputation not to exclude non-responders from the explanatory factor analysis.
In multivariate analysis of DAS score reduction over time, the most important factors predicting a larger DAS reduction were disease severity (DAS) at baseline (i.e. the enrolment visit), no or low steroids (<10 mg prednisolone/day) throughout the study and prevalence of type II diabetes, all p<0.001. Factors predicting poorer response to treatment include the use of NSAIDs, methotrexate and leflunomide anytime during the study (p<0.001) as well as a trend towards previous use of TNFα blockers (N.S.).
The chance of developing a moderate or good EULAR response was significantly lower in patients with a high steroid dose throughout the study vs. those with a low dose (OR 0.76 [0,64; 0.91]), whilst no steroids were not significantly better than low doses (OR 1.05 [0.91; 1.22]). No previous use of biologics also predicted a higher chance of EULAR response (OR 1.24 [1.01; 1.53]).
Conclusions In a study with 458 RA patients treated with anakinra, further post hoc analysis of the raw data suggests that higher disease severity, no or low-dose steroid use or concomitant type II diabetes are predictive of a better response to anakinra treatment. The biological explanations for these associations are unclear. The relation to type II diabetes may be due to common IL-1 involvement in the pathophysiology (ref 2), indicating an innate immune system activation. The relationship to disease severity and steroid use may be related to factors in the pathogenesis and disease course of RA.
Langer H.E. and Missler-Karger B., Int J Clin Pharm Res 2003;23:119-128
Larsen C.M. et al, N Engl J Med 2007;356:1517-26
Disclosure of Interest: B. Missler-Karger Consultant for: Swedish Orphan Biovitrum AB (publ) and Chugai Pharmaceuticals, M. Leu Consultant for: Swedish Orphan Biovitrum AB (publ), M. Raboisson Employee of: Swedish Orphan Biovitrum AB (publ), B. Pilstrom Shareholder of: Swedish Orphan Biovitrum AB (publ), Employee of: Swedish Orphan Biovitrum AB (publ)